MDEV-33710 Implement UUID_TIMESTAMP() function #4496

varundeepsaini · 2025-12-23T19:21:46Z

The Jira issue number for this PR is: MDEV-33710

Description

Adds UUID_TIMESTAMP(uuid) function to extract timestamps from UUIDv1 and UUIDv7 values. Returns NULL for other versions (e.g., v4) or invalid input.

Release Notes

New function: UUID_TIMESTAMP(uuid) - Returns the generation timestamp from UUIDv1/v7 as TIMESTAMP(6).

How can this PR be tested?

./mysql-test/mtr --suite=type_uuid func_uuid_timestamp

Basing the PR against the correct MariaDB version

This is a new feature and the PR is based against the main branch.
This is a bug fix.

PR quality check

I checked the CODING_STANDARDS.md file and my PR conforms to this where appropriate.
For any trivial modifications to the PR, I am ok with the reviewer making the changes themselves.

…v1 and UUIDv7 values

plugin/type_uuid/mysql-test/type_uuid/func_uuid_timestamp.test

plugin/type_uuid/item_uuidfunc.cc

plugin/type_uuid/item_uuidfunc.h

plugin/type_uuid/item_uuidfunc.cc

FooBarrior · 2025-12-24T20:16:40Z

I have some concern about the tests... it's not easy to validate the UUID values used there.
One approach I can think of is:

set timestamp to a fixed value. This will affect NOW() output
See that value returned by UUID_TIMESTAMP(UUID*()) matches our value

We can test a couple of random values, plus corner cases: min and max value for timestamp variable, and for UUID

FooBarrior · 2025-12-24T21:27:29Z

Strangely, there is a different result outpoutted in amd64-debian-11-debug-ps-embedded test run:

--- /home/buildbot/amd64-debian-11-debug-ps-embedded/build/plugin/type_uuid/mysql-test/type_uuid/func_uuid_timestamp.result	2025-12-23 19:37:10.000000000 +0000
+++ /home/buildbot/amd64-debian-11-debug-ps-embedded/build/plugin/type_uuid/mysql-test/type_uuid/func_uuid_timestamp.reject	2025-12-23 19:42:40.539670984 +0000
@@ -1,9 +1,9 @@
 SELECT UUID_TIMESTAMP('018d17f3-e9a3-7000-8000-000000000000');
 UUID_TIMESTAMP('018d17f3-e9a3-7000-8000-000000000000')
-2024-01-17 20:34:37.539000
+2024-01-17 15:04:37.539000
 SELECT UUID_TIMESTAMP('0188b5b8-8000-7000-8000-000000000000');
 UUID_TIMESTAMP('0188b5b8-8000-7000-8000-000000000000')
-2023-06-13 22:35:47.520000
+2023-06-13 17:05:47.520000

Please take a look

varundeepsaini · 2025-12-25T15:13:45Z

i think the test failing is a timezone issue (exact 5.30 HRS), ill fix that

FooBarrior

Thank you @varundeepsaini for the update, and besides such aquick one!
The PR moves in the right direction. Still some work to be done.

mysys/my_uuid.c

plugin/type_uuid/mysql-test/type_uuid/func_uuid_timestamp.result

plugin/type_uuid/item_uuidfunc.cc

FooBarrior · 2025-12-25T18:55:31Z

plugin/type_uuid/item_uuidfunc.cc

+
+  my_time_t seconds;
+  ulong usec;
+  const uchar *buf= (const uchar *) uuid.to_lex_cstring().str;


No functions really need uchar (except the newly added one, which we can change), also uuid is natively stored in char, so why make a conversion?

the mi_ prefix functions use uchar (they do cast it themselves tho, i can remove it from here but again, the cast is still there)

what do you mean? where??

https://github.com/varundeepsaini/server/blob/32e6f8ff2eb54f029e9fced9b17b780e7b7bd6d9/include/myisampack.h#L47-L51

FooBarrior · 2025-12-25T18:58:25Z

plugin/type_uuid/mysql-test/type_uuid/func_uuid_timestamp.test

+SELECT UUID_TIMESTAMP('1ee81f00-a8a8-11ee-8000-000000000000') IS NOT NULL AS has_timestamp;
+
+# Test dynamic UUID generation - timestamp should match NOW()
+SELECT TIMESTAMPDIFF(SECOND, UUID_TIMESTAMP(UUID()), NOW()) AS v1_diff_seconds;


This test is (sort of, implicitly) indeterministic: each time it will check against the new NOW value. So if it will ever fail (because of the bug), noone will be able to reproduce it.

Fix NOW() with set timestamp statement.

UUID generation uses my_interval_timer() which gets the real system time, not the session's SET timestamp

i looked at other tests as well and tey also use this pattern itself (with NOW() / CURDATE() )

Then using it vs NOW etc would mean test would fail sporadically. Can you show the example?

Besides we had no timestamp extraction before, so what do you mean

i grepped NOW() / CURDATE() and found these tests

https://github.com/varundeepsaini/server/blob/6a3fbfdb2d048a0a1b5762f8a56df35ba6986cf0/mysql-test/suite/innodb/t/update_time.test#L20-L25

https://github.com/varundeepsaini/server/blob/36a2a185fe18d31a644da46cfabd9757a379280c/mysql-test/suite/galera/t/MW-292.test#L76-L78

hence did something like that

We can go the other way:
Generate some UUID, somewhere, for example https://www.uuidgenerator.net/, or right in mariadb
Decode: https://www.uuidtools.com/decode
For example, 2731d872-e26e-11f0-8de9-0242ac120002 has 2025-12-26 15:18:38.034341.0 UTC
add test that outputs both correct timestamp (from decoder) and the result from your function

Alternatively, we could add a mockup for using timestamp in my_uuid, but maybe even better if we will do it in the less invasive way.

mysys/my_uuid.c

FooBarrior · 2025-12-25T19:03:23Z

mysys/my_uuid.c

+    Bytes 6-7:  version (4 bits) + sub-millisecond precision (12 bits)
+*/
+
+my_bool my_uuid_extract_ts(const uchar *uuid, my_time_t *seconds, ulong *usec)


Let's use just int.
Or alternatively, bool, with true meaning success

plugin/type_uuid/item_uuidfunc.cc

FooBarrior · 2025-12-25T19:06:44Z

Never mind the failing appveyor, it is used to be failing. However it may be useful to check the results, as it can reveal some windows build quirks (for example msvc doesn't like bool vs int mismatch, and has a 32-bit long )

plugin/type_uuid/mysql-test/type_uuid/func_uuid_timestamp.test

Implement UUID_TIMESTAMP() function that extracts timestamp from UUID…

9aa596f

…v1 and UUIDv7 values

varundeepsaini force-pushed the MDEV-33710-uuid-timestamp branch from 5bbed91 to 9aa596f Compare December 23, 2025 19:33

FooBarrior requested changes Dec 24, 2025

View reviewed changes

refactoring

557584c

FooBarrior requested changes Dec 25, 2025

View reviewed changes

plugin/type_uuid/item_uuidfunc.cc Outdated Show resolved Hide resolved

refactoring

676675f

FooBarrior requested review from FooBarrior and removed request for FooBarrior December 26, 2025 14:04

FooBarrior requested changes Dec 26, 2025

View reviewed changes

plugin/type_uuid/mysql-test/type_uuid/func_uuid_timestamp.test Show resolved Hide resolved

plugin/type_uuid/mysql-test/type_uuid/func_uuid_timestamp.test Show resolved Hide resolved

fixed test

75c06e7

varundeepsaini requested a review from FooBarrior December 26, 2025 17:08

Uh oh!

MDEV-33710 Implement UUID_TIMESTAMP() function #4496

Are you sure you want to change the base?

MDEV-33710 Implement UUID_TIMESTAMP() function #4496

Conversation

varundeepsaini commented Dec 23, 2025

Description

Release Notes

How can this PR be tested?

Basing the PR against the correct MariaDB version

PR quality check

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

FooBarrior commented Dec 24, 2025

Uh oh!

FooBarrior commented Dec 24, 2025

Uh oh!

varundeepsaini commented Dec 25, 2025

Uh oh!

FooBarrior left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

varundeepsaini Dec 25, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

FooBarrior commented Dec 25, 2025

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

2 participants

FooBarrior left a comment •

edited

Loading

varundeepsaini Dec 25, 2025 •

edited

Loading